A subset of ipRGCs regulates both maturation of the circadian clock and segregation of retinogeniculate projections in mice

We utilized two allelic changes at the melanopsin locus to ablate ipRGCs at different ages. As we have published previously, expression of attenuated diphtheria toxin (aDTA) from the melanopsin (Opn4) locus (Opn4^aDTA) results in ablation of mostly the M1 ipRGCs and does so only at adult ages (Güler et al., 2008) (Figure 1A,B,D,E). A newly generated line, expresses the full-strength version of diphtheria toxin (DTA) (Opn4^DTA) and causes ablation of ipRGCs at early postnatal stages (Figure 1; Figure 1—figure supplement 1A–B; Supplementary file 1). To quantify ipRGC loss in the aDTA and DTA lines, we used two genetic labeling methods. The Opn4^Cre; Z/AP mice, in which alkaline phosphatase (AP) expression is dependent on Cre expression in ipRGCs, labels all subtypes of ipRGCs (Ecker et al., 2010). Whereas in Opn4^LacZ mice only M1 ipRGCs are labeled following X-Gal staining for β-galactosidase activity (Hattar et al., 2002; McNeill et al., 2011). In animals heterozygous for DTA and Cre (Opn4^Cre/DTA; Z/AP mice; Supplementary file 1), ipRGCs were reduced in number at birth (Figure 1A,C). Total ipRGC number declined until P14, when approximately 500 cells survived, and then remained constant thereafter through 1year of age (Figure 1A,C and E). Using the LacZ locus with the aDTA or DTA loci, we show that at 6 months of age, about 75 M1 ipRGCs survived in Opn4^LacZ/DTA mice (Supplementary file 1), and consistent with our previous report, about 125 M1 ipRGCs survived in Opn4^LacZ/aDTA mice (Güler et al., 2008) (Figure 1F; Supplementary file 1). These results show that even some ipRGCs that express high levels of melanopsin (M1s) can survive the presence of a single dose of the full strength DTA.

We then sought to determine whether two alleles of DTA would result in a more complete ipRGC ablation. Since melanopsin is the only known marker for all ipRGCs, either an allele of Cre or LacZ is required at the melanopsin locus in order for ipRGCs to be labeled and quantified. Neither an anti-melanopsin antibody nor in situ hybridization with a melanopsin probe can be utilized because Cre, LacZ, aDTA, and DTA replace the melanopsin gene and thus result in a knockout for melanopsin. However, retinal innervation of the SCN originates exclusively from ipRGCs and thus the extent of innervation can be used as a proxy for ipRGC loss. To determine whether two copies of DTA (Opn4^DTA/DTA; Supplementary file 1) would produce further ablation of ipRGCs, we therefore examined retinal innervation of the SCN as an indirect measure of ipRGC loss. We labeled all retinal projections by injecting fluorescently labeled cholera toxin subunitβ (CTB, a neuronal tracer) into the eyes and compared SCN innervation in 6-month-old wild-type, heterozygous (Opn4^DTA/+), homozygous attenuated-DTA (Opn4^aDTA/aDTA), and homozygous DTA (Opn4^DTA/DTA) mice (Figure 1D; Supplementary file 1). Sparse retinal fibers were found in the SCN of heterozygous DTA mice (Opn4^DTA/+) as observed in Opn4^Cre/DTA; Z/AP mice (Figure 1A) but these retinal fibers were entirely absent in Opn4^DTA/DTA mice (Figure 1D) suggesting that Opn4^DTA/DTA mice have more extensive ipRGC loss than Opn4^DTA/+ mice. The more substantial and earlier ipRGC loss in Opn4^DTA/DTA mice was also confirmed by examining SCN innervation using CTB injections in P7 wild-type, Opn4^aDTA/aDTA, and Opn4^DTA/DTA mice (Figure 1B).

Given the strength of DTA as a toxin, we assessed Opn4^DTA/DTA mice for off-target effects. General retinal structure was evaluated by staining retinal sections from wild-type and Opn4^DTA/DTA mice with hematoxylin and eosin stains and additionally fluorescent staining for various retinal cell types, including cones, ON bipolar cells, calretinin-positive amacrine and ganglion cells, and Brn3a positive ganglion cells (Figure 1—figure supplement 1C,E–F). Wild-type and Opn4^DTA/DTA mice appeared similar by these stains, and quantification of total retinal thickness, as well as the thickness of each of the retinal layers, based on hematoxylin and eosin staining showed no difference between wild-type and Opn4^DTA/DTA mice (Figure 1—figure supplement 1D). We also counted the number of SMI-32-positive ganglion cells, Brn3a-positive ganglion cells, cones, and starburst amacrine cells (Figure 1—figure supplement 1G–J). When we compared wild-type and Opn4^DTA/DTA mice, there was no difference in cell number of any of these cell types. It is relevant to mention that about 50% of SMI-32 positive ganglion cells are actually M4 ipRGCs (Estevez et al., 2012; Schmidt et al., 2014), which express lower levels of melanopsin compared to M1 ipRGCs (Ecker et al., 2010; Sexton et al., 2015). Thus, expression of two copies of DTA from the melanopsin locus does not even succeed in killing all ipRGCs. Nonetheless, since we lack specific molecular markers for the other non-M1 ipRGC subtypes, we have no means of knowing the exact time course and extent of ablation for each of those subtypes. To assess potential off target effects in relevant brain regions, we counted the total number of nuclei (based on DAPI staining) and the total number of neurons (identified by antibody staining for Hu) in the SCN of wild-type and Opn4^DTA/DTA mice. There was no difference between wild-type and Opn4^DTA/DTA mice in the number of nuclei or neurons when compared in total or section by section through the SCN (Figure 1—figure supplement 2). Furthermore, the total size of the dLGN was also not different between wild-type and Opn4^DTA/DTA mice or among any other tested mouse lines (Figure 4—figure supplement 4E). These data combined demonstrate that DTA efficiently and selectively ablated ipRGCs at early postnatal ages.

We recorded wheel-running activity under a 12:12-LD cycle, to measure circadian photoentrainment, and under constant darkness, to measure the intrinsic properties of the circadian clock (Figure 2A–C). We assessed wild-type and Opn4^LacZ/LacZ (a null allele for melanopsin) mice as controls and also recorded wheel-running activity of Opn4^aDTA/aDTA, Opn4^DTA/DTA, Opn4^DTA/+, and Opn4^DTA/LacZ mice (Figure 2A; Figure 2—figure supplement 1; Supplementary file 1). Opn4^LacZ/LacZ and wild-type mice showed normal photoentrainment, phase shifting (at CT16), and free-running periods (Figure 2A–C). As previously published, Opn4^aDTA/aDTA mice were unable to photoentrain or phase shift and free-ran under all lighting conditions with a period comparable to controls (Güler et al., 2008) (Figure 2A–C). Opn4^DTA/+ mice photoentrained, phase shifted (at CT16), and free-ran in constant darkness with a normal circadian period (Figure 2A–C) indicating that residual retinal input to the SCN in heterozygotes is sufficient for circadian photoentrainment. In the Opn4^DTA/LacZ mice, the ~500 remaining ipRGCs (~75 M1 ipRGCs) in heterozygous DTA animals, further lose their intrinsic photoresponses due to the loss of both melanopsin alleles and thus can only relay rod/cone input to the brain. In these mice, circadian responses to light were highly attenuated, with inconsistent entrainment and phase-shifting, and exhibited a free-running period comparable to control mice (Figure 2A–C, Figure 2—figure supplement 1A). Similar to Opn4^aDTA/aDTA mice, Opn4^DTA/DTA mice did not photoentrain; however, they free-ran with an abnormally lengthened period (Figure 2A,B; Figure 2—figure supplement 1B). Because full-strength DTA, but not aDTA, kills ipRGCs during early development, these findings suggested the hypothesis that ipRGCs act during early postnatal ages to establish a normal circadian period length. We performed a general examination of expression of transcription factors important for SCN development and a neuropeptide relevant for clock function to examine the SCN, and found no apparent disruption in the SCN of Opn4^DTA/DTA mice (Figure 2—figure supplement 2). Thus, loss of ipRGC innervation does not drastically alter SCN development, and it is possible the lengthened period stems from a loss of direct ipRGC-dependent regulation of SCN neurons. Alternatively, this effect could be due to altered input from the intergeniculate leaflet (IGL), which also receives innervation by ipRGCs.

To assess the relevance of early ablation of ipRGCs in causing the lengthened circadian period of Opn4^DTA/DTA mice, we removed both eyes from mice at either P0 or P60 to mimic the killing of ipRGCs early (Opn4^DTA/DTA) versus late (Opn4^aDTA/aDTA). Starting at P74, we recorded wheel-running activity under a 12:12-LD cycle and constant darkness. Since the mice lack eyes, they lacked photic effects on the clock (Figure 2D,E). However, mice enucleated at P0 exhibited a lengthened circadian period similar to Opn4^DTA/DTA mice, whereas mice enucleated at P60 phenocopied Opn4^aDTA/aDTA mice exhibiting a period comparable to intact mice (Figure 2A–B,D–E). Circadian period length has been shown to be stable in adult mice kept under constant darkness for 2 months (Campuzano et al., 1999). Thus, these data corroborate results from Opn4^DTA/DTA mice indicating that ipRGCs function postnatally to set the length of the circadian period.

We asked whether light-driven input from ipRGCs is important for establishing circadian period length. Wild-type animals were raised under either constant darkness or a 12:12-LD cycle. At P60, we assessed the intrinsic circadian period of these animals by recording their wheel-running activity in constant darkness (Figure 3A,B; Figure 3—figure supplement 1). While the phenotype was less penetrant than that the Opn4^DTA/DTA mice and P0 enucleates, most (9 of 16) dark-reared mice exhibited a longer intrinsic circadian period than mice raised in a 12:12-LD cycle and this was stable for duration of our recordings (up to 60 days; Figure 3A,B; Figure 3—figure supplement 1). The lengthened period of the dark-reared animals was comparable to that exhibited by mice enucleated at P0 and mice with early postnatal ablation of ipRGCs (Opn4^DTA/DTA) (Figures 2 and 3; Figure 3—figure supplement 1). It has been thought that the mammalian circadian period length was established independent of sensory input, but our data show that light contributes to this process. In zebrafish, it is known that light is required to initiate expression of clock genes and for generation of rhythms (Ben-Moshe et al., 2014; Kazimi and Cahill, 1999; Ziv et al., 2005); however, this is the first evidence to indicate that light is also required for mammalian clock maturation.

Upon exposure to a 12:12-LD-cycle, dark-reared mice photoentrained, and light exposure during photoentrainment was sufficient to normalize the circadian period length (Figure 3A,B; Figure 3—figure supplement 1). Furthermore, a single short light pulse (3 hr), which did not cause photoentrainment, (Figure 3C,D) was also sufficient to set the circadian period of dark-reared animals (Figure 3C,D). These data indicate that brief light exposure is sufficient to set the circadian period and can even occur in adults, thus, indicating a lack of a critical developmental window for this process.

By P7, ipRGCs have reached their central targets in the brain (Figure 1A–B) (McNeill et al., 2011) and a subset of ipRGCs forms intra-retinal collateral axons that terminate in the inner plexiform layer, where they synapse onto dopaminergic amacrine cells (Joo et al., 2013; Prigge et al., 2016) (Figure 4—figure supplement 1). P7 is in the middle of a critical developmental period for the image-forming visual system. From P0-P14, the retinotopic map is being refined by spontaneously generated neural activity in the retina, termed retinal waves (Ackman et al., 2012; Feller, 2009; Feller et al., 1996; McLaughlin et al., 2003; Meister et al., 1991; Ramoa et al., 1989; Shatz, 1990; Stellwagen and Shatz, 2002; Xu et al. 2011; Zhang et al., 2011). The presence of intra-retinal collaterals on ipRGCs at early postnatal ages provides anatomical means for ipRGCs to influence the retina during an important developmental window. The Opn4^DTA/DTA mice allowed us the opportunity to directly test the involvement of ipRGCs, not just melanopsin-based light sensitivity, in the development and refinement of the image-forming visual system. To investigate this, we traced the axonal projections of retinal ganglion cells in adult wild-type and Opn4^DTA/DTA mice, and observed that, while retinal innervation of the superior colliculus in Opn4^DTA/DTA mice was comparable to wild-type (Figure 4—figure supplement 2), eye-specific axonal segregation in the dLGN was severely disrupted (Figure 4A,B; Figure 4—figure supplements 3 and 4). In Opn4^DTA/DTA mice, the ipsilateral zone had poorly defined boundaries and substantial contralateral innervation (Figure 4A). We quantified eye-specific axonal segregation in the dLGN using two distinct methods (Datwani et al., 2009; Demas et al., 2006; Renna et al., 2011) (Figure 4B; Figure 4—figure supplement 3). In Opn4^DTA/DTA mice, there was a significant increase in the percentage of pixels with overlapping input from the two eyes, and this eye-specific segregation defect was most severe in the caudal dLGN (Figure 4A, Figure 4—figure supplement 4F). The total amount of ipsilateral and contralateral input and total dLGN size were similar across all tested genotypes (Figure 4C,D, Figure 4—figure supplement 4C–E). Abnormal eye-specific segregation was not observed in either melanopsin knockout mice (Opn4^LacZ/LacZ) or mice with ipRGCs ablated during adulthood (Opn4^aDTA/aDTA) (Figure 4—figure supplement 4A–D). Importantly, Opn4^DTA/+ mice, which have substantial ipRGC loss, although to lesser degree than Opn4^DTA/DTA mice, exhibited disrupted eye-specific segregation that was intermediate between wild-type and Opn4^DTA/DTA mice (Figure 4A,B, Figure 4—figure supplement 4F). Combined, these data reveal an developmental role for ipRGCs in the segregation of all RGC inputs to the dLGN.

To conclusively determine whether ipRGCs function during early postnatal ages to regulate refinement of retinogeniculate projections, we examined eye-specific segregation at P8 and, similar to adult mice, found an increase in overlapping ipsilateral and contralateral projections in Opn4^DTA/DTA mice, but no difference in total retinal input compared to wild-type (Figure 5A–D). These data indicate that ipRGCs function at early postnatal ages to mediate refinement of the image-forming visual system by regulating eye-specific segregation of retinogeniculate projections.

To examine the functional consequence of altered eye-specific segregation, we compared wild-type, Opn4^LacZ/LacZ, and Opn4^DTA/DTA mice in two behavioral tests of visual acuity: the virtual optomotor system and the visual water task (Douglas et al., 2005; Prusky et al., 2000). By both measures, Opn4^DTA/DTA animals exhibited reduced visual acuity compared to controls (wild-type and Opn4^LacZ/LacZ; Figure 4E,F).

Heterozygous DTA animals (Opn4^DTA/+) have less ipRGC loss than homozygous mice (Opn4^DTA/DTA) (Figure 1D; Supplementary file 1) and have intermediate deficits in eye-specific segregation compared to homozygous and wild-type mice (Figure 4A,B; Figure 4—figure supplement 4F). Remarkably, Opn4^DTA/+ mice also exhibited an intermediate reduction in visual acuity (Figure 4E,F). Thus, the severity of the deficits in eye-specific segregation and visual functions were correlated with the extent of ipRGC loss. While many factors including circadian time and the pupillary light reflex contribute to visual acuity, these factors cannot explain the deficits in visual acuity observed in Opn4^DTA/+ and Opn4^DTA/DTA mice. First, the loss of acuity in Opn4^DTA/DTA cannot be explained by a loss of photoentrainment and the possibility that we are testing acuity at different circadian times, because Opn4^DTA/+ mice, which photoentrain also exhibit deficits in visual acuity. In addition, Opn4^aDTA/aDTA mice, which free run, do not exhibit the substantial reduction in acuity observed in Opn4^DTA/DTA mice. Furthermore, loss of the pupillary light reflex cannot fully explain the deficits in visual acuity because when the pupil is fully dilated with atropine there is only a minor reduction in acuity and it is not comparable to the substantial deficits observed in Opn4^DTA/+ and Opn4^DTA/DTA mice (Güler et al., 2008). However, it is important to note that we cannot rule out contributions from direct effects of ipRGCs on retinal functions. For example, the role of M1 ipRGCs in regulating dopamine in the retina, could contribute (Dkhissi-Benyahya et al., 2013; Prigge et al., 2016; Zhang et al., 2012).

Eye-specific segregation deficits in Opn4^DTA/DTA mice were observed by P8, and since light through ipRGCs modulates retinal wave activity and retinal waves drive ipRGC spiking (Renna et al., 2011), we hypothesized that Opn4^DTA/DTA have altered spontaneous retinal activity in darkness. We recorded wave activity at P6 in the dark on a multielectrode array in wild-type and Opn4^DTA/DTA retinas (Figure 5E–H; Supplementary file 2). Spiking properties of RGCs during waves were significantly altered in Opn4^DTA/DTA mice. Wave-associated bursts (WABs) were significantly longer in duration, and had higher firing rates, and shorter inter-burst intervals (Figure 5E,G–H, Supplementary file 2) than they did in wild-type controls. There were also significantly more total spikes as well as more spikes outside of bursts (Figure 5E,G–H, Supplementary file 2). Correlated spiking activity between neurons was very similar between the genotypes, as measured by the spike time tiling coefficient (STTC; [Cutts and Eglen, 2014])(Figure 5F). These data indicate that ipRGCs are critical for normal retinal wave activity, even in darkness, and together with our anatomical findings (Figure 4A–D; Figure 4—figure supplements 3–4), allow us to suggest that ipRGCs mediate eye-specific segregation of RGC projections to the dLGN by regulating the spiking properties of conventional RGCs during retinal waves.

There are currently five identified subtypes of ipRGCs. M1 ipRGCs project to non-image-forming brain centers, while M2-M5 ipRGCs project at least in part to image-forming targets (Ecker et al., 2010; Schmidt et al., 2011; Schmidt and Kofuji, 2009). The M1 subtype can be further subdivided based on expression of the transcription factor Pou4f2 (also referred to as Brn3b), which is also expressed in all non-M1 ipRGCs (Chen et al., 2011).~200 Pou4f2-negative-M1 ipRGCs project exclusively to circadian centers (predominantly to the SCN) (Figure 4—figure supplement 5B) (Chen et al., 2011) and are the subset of ipRGCs that have intra-retinal collateral axons (data not shown). This population of cells are sufficient for circadian photoentrainment in the presence of one copy of melanopsin (Chen et al., 2011). As described previously, Pou4f2-positive M1 ipRGCs and non-M1 ipRGCs can be selectively ablated by crossing Opn4^Cre/+ mice with Pou4f2^Z-DTA/+ mice (previously published as Brn3b^Z-DTA/+ mice), in which a floxed stop cassette followed by DTA was inserted into the Pou4f2 locus (Chen et al., 2011; Mu et al., 2005). In doubly heterozygous offspring (Opn4^Cre/+; Pou4f2^Z-DTA/+, previously published as Opn4^Cre/+;Brn3b^Z-DTA/+ mice, Supplementary file 1), Pou4f2-positive ipRGCs (some M1 and all non-M1 ipRGCs) are ablated by P7, leaving a subset of M1 Pou4f2-negative ipRGCs (~200 cells) and all conventional RGCs (Chen et al., 2011) (Figure 4—figure supplement 5A, Supplementary file 1). In these mice, the SCN remains innervated by ipRGCs, the IGL receives partial innervation, and ipRGC innervation of image-forming brain regions is entirely abolished (Chen et al., 2011) (Figure 4—figure supplement 5B).

In Opn4^Cre/+; Pou4f2^Z-DTA/+ mice, ablation of Pou4f2-positive ipRGCs is complete by P7, and at this age only the ~200 Pou4f2-negative ipRGCs remain (Figure 4—figure supplement 5A; Supplementary file 1). Opn4^Cre/+; Pou4f2^Z-DTA/+ mice can photoentrain and free-run with a circadian period indistinguishable from controls (Chen et al., 2011), indicating that the remaining Pou4f2-negative-M1 ipRGCs are sufficient to set the circadian period. Organization of eye-specific retinogeniculate projections in Opn4^Cre/+; Pou4f2^Z-DTA/+ is also indistinguishable from controls (Figure 4A–D; Figure 4—figure supplement 4A–D). Together these results demonstrate that a single subpopulation of ipRGCs are critically involved in the development of networks devoted to image-forming vision and the circadian clock. This small population of cells is sufficient for setting of the circadian period as well as proper segregation of retinogeniculate circuitry.

Animals were housed and treated in accordance with NIH and IACUC guidelines, and used protocols approved by the Johns Hopkins University and Brown University Animal Care and Use Committees (Protocol numbers MO16A212, and 1010040).

All statistical tests were performed in Graphpad Prism 6, except for retinal wave analysis the detail of which are described below. Specific statistical comparisons are listed in the figure captions.

To generate Opn4^DTA mice, we used the targeting arms and general strategy detailed in (Ecker et al., 2010; Güler et al., 2008; Hattar et al., 2002). The construct contained a 4.4 kb sequence immediately 5’ of the start codon for mouse melanopsin, followed by the coding sequence for diphtheria toxin A (DTA) subunit, an internal ribosomal entry site (IRES), the coding sequence for tauLacZ, and a self-excising neomycin resistance construct (loxP-tAce-Cre-Pol II-Neo-loxP) (Figure 1—figure supplement 1A). Embryonic stem (ES) cells were first screened for homologous recombination by PCR, and then homologous recombination was confirmed with Southern blot analysis (a restriction digestion with SpeI resulted in a 10.7 kb band for the wild-type allele and a 5.3 kb band for the recombined allele) (Figure 1—figure supplement 1B). The blastocyst injection was performed by the Johns Hopkins transgenic core facility. The germline transmission was obtained by crossing chimeric males with C57Bl/6J females. The genotyping was done by PCR. The DTA allele was detected with the primers: AACTTTTCTTCGTACCACGG (forward) and ACTCATACATCGCATCTTGG (reverse), and the wild-type allele was detected with the primers: CCCCTGCTCATCATCATCTTCTG (forward) and TGACAATCAGTGCGACCTTGGC (reverse). Opn4^DTA/+ and Opn4^DTA/DTA mice are viable, fertile, and do not exhibit any gross abnormalities in size.

A cre-mediated alkaline phosphatase (AP) reporter, provided by Tudor Badea in Jeremy Nathan’s lab, was expressed in conjunction with Opn4^Cre (Ecker et al., 2010). Mice were deeply anesthetized with 30 ml/kg Avertin and then intracardially perfused with phosphate-buffered saline for 3 min followed by 40 ml of 4% paraformaldehyde. Brains and retinas were isolated and post-fixed for 40 min in 4% paraformaldehyde. Brains were mounted in 3% agarose and then cut into 200 µm sections on a vibrating microtome (Vibra-tome 1000 Plus). Tissue was heat-inactivated for overnight at 65°C. Alkaline phosphatase histochemistry was performed using NBT/BCIP tablets (Roche) for 2–4 hr in the dark with constant shaking. Tissue was washed three times with phosphate-buffered saline containing 0.1% Tween-20 (Sigma-Aldrich). Retinas were mounted immediately and imaged. Brains were fixed 3 hr in 4% paraformaldehyde at 4°C, then counterstained with 1:5 Fast Red nuclear stain (Vector Laboratories) in water for 7 min. The sections were then dehydrated in an ethanol series, and after at least in hour in 100% ethanol, the sections were cleared in a 2:1 mixture of benzyl benzoate:benzyl alcohol (Sigma-Aldrich), mounted in glycerol, and imaged immediately. To measure cell density, we counted the number of ipRGCs, in four representative areas of each retina, and calculated the density of ipRGCs per mm².

Mice were deeply anesthetized with 30 ml/kg Avertin followed by cervical dislocation. Eyes were isolated and fixed in 4% paraformaldehyde for 10 min. Retinas were dissected out and then incubated in buffer B (100 mM phosphate buffer at pH 7.4, 2 mM MgCl₂, 0.01% sodium deoxycholate, 0.02% IGEPAL) then stained for 3 days in buffer B plus 5 mM potassium ferricyanide, 5 mM potassium ferrocyanide and 1 mg/ml X-gal as described in Hattar et al. (2002).

Animals were anesthetized with 30 ml/kg Avertin, and eyes were removed fixed in 4% PFA for 1 hr. Retinas were dissected in PBS, placed in cartridges (Tissue-Tek Biopsy Uni-Cassette), and processed and embedded in paraffin overnight. Eyecups were sectioned at 6 μm. Resulting sections were deparaffinize by immersion in two changes of xylene for 10 min each. Sections were then rehydrated in descending series of ethanol ending in water for 5 min. Sections were stained with hematoxylin for 30 s, washed with tap water for 10 dips, placed briefly in 0.1% sodium bicarbonate, and then rinsed in clean tap water for 10 dips. Sections were rinsed in 70% ethanol for five dips and stained in eosin for 20 s. Sectioned were dehydrated with an ascending series of ethanol, ending with two washes of 100% ethanol. Sections were placed in two washes of xylene (5 min each), and mounted in Permount.

Whole eyes were fixed for 15 min in 2% paraformaldehyde (PFA) diluted in PBS and were then dissected to remove the cornea and lens. Whole eyecups were fixed for additional 45 min in 2% PFA diluted in PBS. Further dissection was done to release the retinas from the RPE, and four nicks were made so that the retina would lay flat. Whole retinas were blocked in 500 mL of PBS containing 0.3% Triton X100% and 3% goat serum for 2 hr at RT. Either mouse anti-Brn3a (Millipore, cat #AB5945, RRID:AB_92154↗) (1:250), rabbit anti-γ13 (generously provided by Robert Margolskee, RRID: AB_2314434↗)(1:500), rabbit anti-calretinin (Swant, cat #CR 7699/3 hr, RRID: AB_10000321↗)(1:500), goat anti-ChAT (Millipore, cat #Ab144p, RRID: AB_262156↗) (1:200), or mouse anti-SMI-32 (Covance, cat #SMI-32R, RRID: AB_2315331↗) (1:500) was diluted in blocking solution and incubated overnight at 4°C. Retinas were washed 10 min in three changes of PBS, then placed in 1:500 Alexa Fluor secondary antibody (Invitrogen) overnight at 4°C. Retinas were washed as above and mounted flat on slides in VectaShield (Vector Labs, RRID: AB_2336789↗). To measure cell density was measured by counting cells in four representative areas of each retina, and density was calculated as the number of cells per mm².

For SCN cell counts, mice were perfused with cold 4% PFA, brains were dissected out and then cryoprotected in 30% sucrose, frozen in OCT, and 25 μm serial coronal sections containing the SCN were taken. Slides were blocked in PBS containing 0.3% Triton X100 and 3% goat serum for 3 hr at RT and then incubated overnight at 4°C in 1:200 monoclonal mouse immunoglobulin G (IgG) 2b anti-human-HuC/D (Thermofisher, cat# A-21271, RRID: AB_2096358↗). Slides were then washed in PBS and then incubated in 1:500 goat anti-mouse IgG2b (isotype of secondary antibody is critical (Thermofisher, RRID: AB_429670↗)) overnight at 4°C. Slides were then washed in PBS and mounted in VectaShield (Vector Labs, RRID: AB_2336789↗) with DAPI. For counting of DAPI-labeled nuclei and Hu-labeled cell bodies, we used a program that was coded in Mathmatica (Wolfram) and was previously described in Martinelli et al. (2016).

Mice were deeply anesthetized with 30 ml/kg Avertin followed by cervical dislocation. Eyes were isolated and fixed in 4% PFA for 1 hr. Retinas were dissected out and then incubated for 2 hr in Alexa 488-conjugated lectin peanut agglutinin (PNA) (Invitrogen cat# L21409↗, RRID: AB_2315178↗) diluted 1:200 in PBS with 0.3% Triton X100% and 3% goat serum. Retinas were then washed in PBS and mounted in VectaShield (Vector Labs, RRID: AB_2336789↗).

Mice were anesthetized with a ketamine/xylazine mixture before transcardially perfusing with 0.9% saline followed by 4% PFA in PBS, pH 7.4. All tissues were post fixed in 4% PFA overnight at 4°C, cryoprotected in 20% sucrose in PBS, frozen in OCT. Compound Embedding Medium (Tissue-Tek), and stored at −75°C. Serial sections (20 μm) were cut on a Hacker cryostat and thaw mounted on Superfrost Plus slides (Thermo Fisher Scientific). Five adjacent sets of sections were prepared from each postnatal age and stored at –20°C. Probes were generated as described in VanDunk et al. (2011). Slides were immersed in 4% PFA, permeabilized with proteinase K, returned to 4% PFA before being washed in 0.1 M triethanolamine-HCl with 0.25% acetic anhydride. Once blocked in hybridization buffer at 65°C slides were incubated in hybridization buffer containing 1–2 µg/ml DIG-labeled anti-sense cRNA overnight at 65°C. Slides were then washed in 2XSSC buffer at 62°C, washed in 0.2XSSC at 65°C, blocked with 10% normal horse serum (NHS) in 0.1M PBS, and incubated in alkaline phosphatase labeled anti-DIG antibody (1:2000 in 10% NHS; Roche, RRID: AB_514497↗) overnight. Sections were washed and color was visualized using Nitro blue tetrazolium and 5-Bromo-4-chloro-3-indolyl phosphate (Roche). Staining was stopped after visual inspection. Sections were washed, fixed in 4% PFA, and coverslipped in 90% glycerol, Vectashield Mounting Medium (Vector Laboratories, RRID: AB_2336789↗), or UltraCruz Mounting Media with DAPI (Santa Cruz Biotechnology).

Images were acquired using a Nikon Eclipse 90i microscope, Photometrics Coolsnap HQ2 camera with a Prior Scientific ProScan II motorized translation stage, and acquired in Volocity (PerkinElmer Life and Analytical Sciences). Images were exported as 8bit JPEG or TIFF files. All images were adjusted for clarity by filtering and/or modifying levels, as necessary, in Photoshop (Adobe Systems).

In order to remove the eyes, P0 mice were placed on ice for 2 min, and then a 1–2 mm incision was made across each eyelid using a sterile scalpel blade. The scalpel blade was then used to puncture the eyes and forceps were used to pull the eyes free of the orbitals. P60 mice were first anesthetized with intraperitoneal injection of 20 mL/kg of Avertin. Fingers were placed on either side of the eye causing it to bulge, a curved pair of scissors was placed between the eye and the skin, and the optic nerve was cut. Bleeding was controlled by orbital pressure. The animal was monitored over the next several days for signs of infection.

Mice were placed in cages with a 4.5-inch running wheel, and their activity was monitored with VitalView software (Mini Mitter), and cages were changed at least every 2 weeks. All free-running periods and phase shifts were calculated with ClockLab (Actimetrics).

6-month-old wild-type (n = 18 mice), Opn4^LacZ/LacZ (n = 17), Opn4^DTA/+ (n = 8), Opn4^DTA/LacZ (n = 7), Opn4^DTA/DTA (n = 7), and Opn4^aDTA/aDTA (n = 18) mice were placed in 12:12 LD for 10 days followed by constant darkness for 14 days. For phase-shifting experiments, a subset of wild-type, Opn4^LacZ/LacZ, Opn4^aDTA/aDTA were used. Wild-type (n = 7), Opn4^LacZ/LacZ (n = 8), Opn4^DTA/+ (n = 8), Opn4^DTA/LacZ (n = 7), Opn4^DTA/DTA (n = 7), and Opn4^aDTA/aDTA (n = 8) mice were exposed to a light pulse (500 lx; CT16) for 15 min, after being in constant dark for 14 days.

P0 and P60 ennucleated animals were placed in 12:12 LD for 24 days followed by constant darkness for 14 days.

Dark-reared animals and control mice (raised in 12:12 LD) were placed in constant darkness without any exposure to light. Wheel running behavior was recorded in constant darkness for 1–3 months. Mice were then either given a 3-hr light pulse (LD reared: n = 17; DD reared: n = 12) or placed in 12:12 LD for 14 days and then presented with a 6-hr shift and allowed to re-entrain for 14 days (LD reared: n = 13; DD reared: n = 16). Mice were then returned to constant darkness for 1 month.

WT (n = 12), Opn4^LacZ/LacZ (n = 6), Opn4^aDTA/aDTA (n = 6), Opn4^DTA/+ (n = 8), Opn4^DTA/DTA (n = 10), and Opn4^Cre/+ (n = 8), and Opn4^Cre/+; Pou4f2^Z-DTA/+ (n = 6) mice used for examination of adult central projections were raised in a standard 12:12 LD cycle. 3-6-month-old mice were anaesthetized with 20 mL/kg of Avertin. Eyes were injected intravitreally using a glass pipet with approximately 2 μl of cholera toxin B subunit conjugated with Alexa Fluor 488 (Thermofisher, cat# C34775↗) or Alexa Fluor 594 (Thermofisher, cat# C22842↗) using a Harvard Apparatus HL-190 picospritzer. CTB-488 was used at a concentration of 6.25 µg/µL and CTB-594 was used at 5 µg/µL for all injections. Three days after injection, mice were perfused with 4% PFA, and brains were isolated, cryoprotected in 30% sucrose, frozen in OCT, and 40 µm sections were taken using a cryostat. Sections were dried overnight, mounted in VectaShield, and imaged on a Zeiss Imager M1 upright epifluorescence microscope (AxioVision). Retinas were also dissected, mounted in vectashield, and examined for good injection quality.

Mice used to examine LGN projections at P8 were born and raised in 12:12-LD cycle. At P7, mice were anaesthetized on ice, and ocular injections of contrasting fluorescent anterograde tracers (cholera toxin subunit B (CTB)-Alexa-488 and CTB-Alexa-594) were made. At P8, mice were perfused with 4% PFA, and brains were isolated, cryoprotected in 30% sucrose, frozen in OCT, and 40 µm sections were taken using a cryostat. Sections were dried overnight, mounted in VectaShield (Vector Laboratories, RRID: AB_2336789↗), and imaged on a Zeiss Imager M1 upright epifluorescence microscope (AxioVision). Retinas were also dissected, mounted in vectashield, and examined for good injection quality. Only animals with complete retinal labeling were assessed further. For assessment of SCN innervation at P7, mice were injected with CTB-Alexa-488 at P4. At P7, mice were perfused with 4% PFA, and brains were isolated, cryoprotected in 30% sucrose, frozen in OCT, and 40 µm sections were taken using a cryostat. Sections were dried overnight, mounted in VectaShield (Vector Laboratories, RRID: AB_2336789↗), and imaged on Zeiss LSM 700 Confocal.

Quantifications were performed with the analyzer blinded to the genotypes being measured.

Similar to (Fernandez et al., 2012), Coronal sections were used for the SC reconstruction using Matlab (Math Work). For each section, the retino-recipient SC was outlined, and the total retinotopic area was calculated. Images were converted to 8-bits of grey scale and the optic density of CTB-staining was calculated. The total length was measured and divided in bins (4 µm) from the medial to lateral region. The CTB density was obtained by dividing the total pixel area by CTB+ pixels. Finally, a colorimetric thermal representation was applied (from 0% = blue to 100% = red). All sections containing the SC were used for a final reconstruction of the retinal projection to the SC.

Mice were shipped overnight at postnatal day 5 (Opn4^DTA/DTA from Johns Hopkins University and wild-type control mice from Jackson Laboratories). All procedures involving the use of animals were in accordance with the National Institutes of Health and approved by the Brown University Institutional Animal Care and Use Committee. Upon arrival at Brown University, the P6 mice were sacrificed via a lethal intraperitoneal injection of Beuthanasia. As previously described, retinas were extracted and placed ganglion-cell side down on the array under dim red light (Renna et al., 2011). The retinas were continuously superfused with oxygenated Ames solution maintained at 36–37°C, and were kept in constant darkness. Recordings were made after at least 30 min of dark adaptation.

The raw analog data were digitized using MC Rack software (Multi Channel Systems) before being filtered. Further processing was done in OfflineSorter (Plexon Inc.). In order to determine the number of neurons being sampled by each electrode, we filtered the raw digitized data using a 125 Hz high-pass filter. A threshold of 5 standard deviations from the baseline voltage was used as the criterion for a spike. The candidate spikes were then sorted using OfflineSorter (Plexon Inc.) for the first two principal components of the waveforms, using a standard T-distribution E-M algorithm (Shoham et al., 2003). Spikes appearing within a 1 ms window on 70% of channels were assumed to be artifacts (caused by bubbles or other disturbances) and were discarded.

Using the virtual optokinetic system (Douglas et al., 2005), we placed individual mice in a box created using four computer screens, which display sine wave gratings that move to create virtual cylinder. If mice can see the gratings, they track the moving bars by turning their head. To obtain an estimate of visual acuity using this setup, we increased the spatial frequency (SF) of the gratings until the mice no longer tracked the movement of the gratings.

In the visual water task (Prusky et al., 2000), mice are place in a trapezoidal shaped maze that contains water and has a divider down the middle to create two arms. At the end of one arm, a gray panel in displayed and in the other arm, sine wave grating is displayed. Mice were trained, using gratings with a low SF, to associate the gratings with a hidden platform that allows them to escape from the water. The location of the grey panel and gratings with the hidden platform were moved between the arms in a pseudorandom pattern that mice cannot memorize. Any entrances into the arm containing the grey panel were recorded as incorrect, after the mice could reliably swim to the platform with greater that 90% accuracy, the SF of the gratings was increased. An animal’s threshold was considered to have been reached once they failed to have better than 70% accuracy out of 10 trials. This threshold was confirmed by retesting the previous SF, which is 0.02cycles/degree lower, and then retesting the spatial frequency that the mice initially failed. If the mouse’s behavior repeated where they could see the lower SF, but not one increment higher, then the visual acuity of that mouse was recorded as the last spatial frequency where they had better than 70% accuracy.